New treatment agents against osteoporosis may not only lead to an improved trabecular structure, but can probably also strengthen the cortex. To assess this treatment effects by means of QCT, one has to deal with significant partial volume effects observing this very thin, but compact structure. A new method for cortical thickness estimation has been developed using an analytical deconvolution approach. After estimating the point-spread-function (PSF) of a scanner/kernel combination, the associated line-spread-function (LSF) and edge-spread function (ESF) have been derived. Moreover, using sophisticated 3D segmentation, a radial bone mineral density (BMD) distribution was calculated. By fitting a combined LSF/ESF functional to these distributions assuming full mineralization (1100 mgHA/cc) in a compact cortex structure, we directly get an averaged de-convolved cortical thickness dcCt.Th for the cortex fraction of interest. The spongiosa plateau is reflected by the ESF, whereas the Cortex is translated into the LSF. The nonlinear fit (JMP, SAS Institute, Inc,. USA) directly delivers the spongiosa plateau BMD and the cortical thickness and its relative position to the initial segmentation. We did the HRqQCT processing using software provided by Scanco Medical (Switzerland) for a vertical cortex segmentation and compared the calculated cortical thicknesses with HRQCT derived data using StructuralInsight (in-house, ITK-based). Besides a normal maximum-sphere based variable (Ct.Th), we also calculate a density-weighted variable (wCt.Th). In addition to the already mentioned dcCt.Th another variable dcCt.Th* is introduced respecting the unusual embedding material of the ten analysed human vertebral bones (ethical approved during Eurogiops/Bioasset studies). Bland-Altman plots show mean differences to the HRpQCT based estimation of (1.17±0.24), (0.16±0.08), (0.13±0.03) and (0.09±0.03) mgHA/cc for Ct.Th, wCt.Th, dcCt.Th and dcCt.Th* respectively. In the first two cases, significant slopes of (1.35±0.21) and (0.59±0.18) are observable, but not for the latter two. Comparing the mean values of the estimates directly, we find (0.36±0.07) for the HRqQCT and (1.52±0.27), (0.52±0.13), (0.49±0.09) and (0.46±0.08) mgHA/cc.again for CT.Th, wCt.Th, dcCt.Th and dcCt.Th* respectively. Therefore we use basic image analysis to improve the estimation of cortical thickness despite the partial volume effect.
17 May 2014 - 20 May 2014